Foetal aortic flow velocity waveforms in healthy and hypertensive pregnant women

The study was approved by the ethics committee of Centro Hospitalar do Porto and all subjects provided informed consent [N/ REF.ª 133/10(086-DEFI/126-CES].

Patients were referred to the hospital by their family doctor according to local pregnancy health policies. We studied singleton pregnancies in healthy women or in women with chronic arterial hypertension without known target organ involvement. Women taking folic acid, vitamin or iron supplements were included. Additionally, acetylsalicylic acid, 100 mg per day, was specifically prescribed to all hypertensive women on their first appointment. Women taking other medications were excluded.

In the first appointment, that coincided with the first ultrasound evaluation, they were observed by a senior specialist who reviewed the patient’s history, verified the absence of diabetes and other endocrine disorders, immune disease, renal and structural heart disease, haematological conditions and chronic infections, and measured blood pressure; gestational age (GA) was verified by sonography to be between 11 and 14 weeks. Hypertension (HT) was defined as systolic pressure ≥140 mmHg and/or diastolic pressure ≥90 mmHg, present before pregnancy or the 20th week [19, 25‐27]. The average of two measurements taken after a 4 hours period of rest was calculated.

Women that met the inclusion criteria and agreed to participate in the study were then enrolled in a longitudinal prospective study that included a trimestral ultrasound evaluation and the recommended regular blood tests. Body mass index was determined upon biometrical data collected before the ultrasound evaluation.

Pregnancies were closely monitored for the appearance of abnormal conditions. These included foetal abnormal Doppler indexes in the umbilical and middle cerebral arteries, and foetal growth <10th and >90th percentile growth curves [28, 29], which were a reason for quitting the follow-up.

All deliveries took place at Centro Hospitalar do Porto and all infants were evaluated by a neonatologist.

Doppler flow study of foetal proximal descending aorta (AoF) was performed in the 101 women included in the study, immediately before performing the regular obstetrical ultrasound evaluation at the 1^st (11–14 weeks), 2^nd (19–22 weeks) and 3^rd (28–32 weeks) trimesters. AoF-RI and PI Doppler examinations were made using a Voluson E8 or a Voluson 730 Pro (GE Healthcare Technologies, USA) ultrasound machine, equipped with multifrequency transvaginal and transabdominal transducers and all measurements were performed by an investigator with extensive experience in Doppler ultrasound, in order to avoid inter-observer variability. Ultrasound colour Doppler cineloops of the foetal descending aorta were obtained using a C5-2 curved-array transducer, with center frequency of 5 MHz for grayscale imaging and 2.5 MHz for colour Doppler imaging. The colour Doppler maximum velocity setting was adjusted to high velocities so that blood flows of the great vessels were homogeneous in colour, showing no aliasing. The size of the sample volume was adapted to the vessel diameter in order to cover it entirely. All recordings used for measurements were performed in the absence of foetal movements. The scanning plane was adjusted to obtain an insonation angle <30°. All AoF Doppler measurements were sampled and the signal updated until three similar consecutive waveforms were obtained – Figure 1. Pulsatility index (PI) was used as a measure of impedance of the flow of blood distal to the sampling point and automatically calculated according to the formula PI = s − d mean where s is the peak d is the minimum and the average is the mean maximum Doppler shift frequency over the cardiac cycle. Resistance index (RI) was automatically calculated using the formula RI = s − d s. Smokers were required to abstain from smoking for at least 2 h prior to examination.

AoF assessment is rather simple albeit on occasions, particularly the first trimester of pregnancy, its visualization was sometimes difficult, consequent to frequent dorsoanterior position and ultrasound attenuation caused by the spine. To overcome this difficulty, in the current study, foetal stimulation with abdominal probe and maternal mobilization into lateral decubitus was successfully employed, in agreement with other’s observations [3, 18].

Univariate data analysis comprised appropriate statistical methods: chi-square test or Fisher’s test (as adequate) for the study of independence amongst two factors, t-test for the difference of means in two independent populations, and one-way analysis of variance for repeated measurements with Tukey’s multiple comparisons test for the assessment of significant statistical differences across more than two means in paired populations.

Multiple linear regression models with errors, that were allowed to be correlated and/or to have unequal variances, were fitted using generalized least squares. Multivariate regression had to be considered due to the experiment’s nature: two different indexes were read on the same set of individuals, once at each trimester of the pregnancy. We looked for adequate global models and compared curves (as functions of time but adjusted for potential confounders) instead of comparing mean indexes between different time points. Random effects models were not used as the concern was on serial correlation and results were to be interpreted at the population level.

The response variable read for index d (RI or PI), in a subject with hypertensive status h (hypertensive or normotensive), at (continuous) time t was denoted by R(d,h,t). Dummy variables had to be considered for the categorical variables index and hypertension status; reference categories were taken to be the resistance index and the normotensive status, respectively. The fitted model was

with residuals ϵ following a normal distribution with mean zero and with a variance-covariance matrix that allowed for an intra-individual time autocorrelation structure of order 1 and for different variances across the indexes and the hypertensive status. Equivalently, the predicted values were given by the non-linear equation

Final regression models were chosen on the basis of the lowest BIC (Bayesian Information Criterion). Residuals normality was improved after the removal of the observations with Pearson residual greater than 3.3.

Intra-observer reliability was obtained from two readings, at the beginning and the end of the scan, on the first 60 recordings of RI and PI in the AoF. This concept concerns the degree to which measurements taken by the observer are consistent. In the present context, there was only one observer, and he was totally unaware of any of the results. Intraclass correlation coefficients (ICC) and 95% confidence intervals were calculated using a two-way mixed-effects model with absolute agreement. The reliability coefficient, which is the difference value that will be exceeded by only 5% of pairs of measurements on the same subject, was calculated as 1.96 times the standard deviation of the difference between pairs of repeated measurements.

All statistical analyses were carried out using the R language and software environment for statistical computation, version 2.12.1 [30]. The significance level was fixed at 0.05.

The unadjusted effect of the gestational trimesters on the mean values of the indexes is merely indicative, and multivariate analyses adjusted for potential confounders and taking the experience design into consideration had to be considered. As the difference between the average evaluation time in trimester 2 and that in trimester 1was approximately equal to the difference between the average evaluation time in trimester 3 and that in trimester 2 (more precisely, the latter is 0.95 times the former), the multivariate regression model considered the variable gestational trimester as continuous. The model also included the index and the hypertensive status as explanatory variables, as described above in the Section Statistical Analysis; its estimated coefficients and respective 95% confidence intervals are presented in Table 3. Known confounding variables such as age, smoking habits, body mass index and number of previous pregnancies were also taken into account in the analysis; as they were not shown to be statistically significant, they were not considered in the final model that allowed the construction/prediction of AoF-RI and AoF-PI curves – Figure 3. Log-transformations and removal of outliers were considered in order to improve the model. Seven observations (all from normotensive women and all regarding PI values, although corresponding to different time points) were identified as having Pearson residuals greater than 3.3. As its removal improved residuals normality, they were eliminated from the model fitting.

At baseline, in the first trimester and for RI only, a significant difference between its mean values in hypertensive and normotensive women was identified - Table 3. However, as gestational time progressed, that effect ceased its significance and the obtained mean values for hypertensive and normotensive women were no longer significantly different from each other. On the other hand, for PI, the scenario was the opposite: although no significant differences between hypertensive and normotensive women were detected at baseline, the mean values of the two groups at the second and third trimesters differed significantly from each other (Table 3).

The model equation written in the original variables had multiplicative effects, implying that coefficients estimates had to be interpreted accordingly. For example, for a fixed index, a fixed hypertensive status and a fixed time point t, the response value obtained for time kt is k ^{β 1} times the response value obtained for t, for any real number k. Nonetheless, for any fixed index and hypertensive status, a significant effect of time was observed. All curves obtained in Figure 3 are significantly different from a flat line therefore reflecting an increase in RI and PI values along the gestation. This follows directly from the statistical significance of all terms containing log(Time) in the considered model (Table 3).

(1) The study was conducted in a sample of uneventful pregnancies. Therefore, it is not possible to establish causal relationships to the development of severe disorders or appearance of serious foetal events that future studies with specific risk groups may provide. (2) Blood pressure was not measured at the moment of each Doppler study; (3) Starting at the first trimester, HT patients were medicated daily with low-dose acetylsalicylic acid, whose contribution can’t be formally excluded.